Upper limb motor pre-clinical assessment in Parkinson's disease using machine learning

Area of Science:

• Biomedical Engineering
• Neuroscience
• Medical Technology

Background:

• Parkinson's disease (PD) is a neurodegenerative disorder with significant motor and non-motor symptoms.
• Idiopathic hyposmia (reduced sense of smell) is a preclinical marker present in over 95% of PD patients.

Purpose of the Study:

• To develop and validate a wearable inertial device (SensHand V1) for objective Parkinson's disease diagnosis.
• To assess the efficacy of motion analysis combined with olfactory screening for early PD risk identification.

Main Methods:

• Collected upper limb motion data using the SensHand V1 wearable device from 30 healthy subjects, 30 individuals with idiopathic hyposmia, and 30 PD patients.
• Computed 48 spatiotemporal and frequency parameters per side and selected a significant feature array for classification.
• Compared three supervised learning algorithms: Support Vector Machine (SVM), Random Forest (RF), and Naïve Bayes.

Main Results:

• Achieved excellent classification accuracy for healthy vs. PD patients (F-Measure 0.95-0.97).
• Demonstrated good performance in a three-group classification including hyposmia patients (0.79 accuracy, 0.80 precision with RF).
• Identified Random Forest classifiers as the optimal approach for this application.

Conclusions:

• The SensHand V1 system shows suitability for objective PD diagnosis.
• A two-step non-invasive procedure combining motion analysis and olfactory screening can aid in identifying individuals at risk for PD.
• This approach helps clinicians detect subtle motor changes indicative of PD onset.